A geometric-sensitivity-difference based algorithm improves object depth-localization for diffuse optical tomography in a circular-array outward-imaging geometry

Abstract

Listen

Translate article

To improve object depth-localization for diffuse optical tomography (DOT) in a circular-array outward-imaging geometry that is subjected to strong sensitivity variation with respect to imaging depth. The authors introduce an alternative DOT image reconstruction approach that optimizes the data-model fit based on the paired measurements corresponding to two pairs of source-detector that share either the source or the detector, in comparison to the conventional method that optimizes the data-model fit based on the unpaired measurements corresponding to individual pairs of source-detector. This alternative approach, namely, geometric-sensitivity-difference (GSD) method, effectively reduces the variation of the reconstruction sensitivity with respect to imaging depth. The DOT image reconstruction based on GSD-scheme applied to same-source source-detector pairs is demonstrated using simulated and experimental continuous-wave measurements in a circular-array outward-imaging geometry, of which the native sensitivity varies strongly with respect to the depth. The outcomes of GSD-based image reconstruction are compared to those of two other methods: one is the conventional baseline method that utilizes the native sensitivity but does not involve depth-compensating scheme; and the other is a reference-compensation approach that employs active and depth-adapted compensation scheme to counteract the dependence of the reconstruction sensitivity with respect to imaging depth. The GSD method generally outperforms the other two methods in localizing the depth of single object, resolving two objects that are azimuthally separated, and estimating the optical property of single object or azimuthally separated dual objects. The GSD method, however, demands more computations due to an increase of the element size of the resulted sensitivity matrix and more matrix multiplications. The GSD method improves the depth localization in the circular-array outward-imaging geometry, by taking advantage of the paired measurements of two source-sharing source-detector-pairs to passively and effectively homogenize the sensitivity of the reconstruction with respect to imaging depth.